JOURNAL OF MEDICAL INTERNET RESEARCH Xue et al Original Paper The Hidden Pandemic of Family Violence During COVID-19: Unsupervised Learning of Tweets Jia Xue1,2, PhD; Junxiang Chen3, PhD; Chen Chen4, PhD; Ran Hu1, MA, MSW; Tingshao Zhu5, PhD 1Factor-Inwentash Faculty of Social Work, University of Toronto, Toronto, ON, Canada 2Faculty of Information, University of Toronto, Toronto, ON, Canada 3School of Medicine, University of Pittsburgh, Pittsburgh, PA, United States 4Middleware System Research Group, University of Toronto, Toronto, ON, Canada 5Institute of Psychology, Chinese Academy of Sciences, Beijing, China Corresponding Author: Jia Xue, PhD Factor-Inwentash Faculty of Social Work University of Toronto 246 Bloor St W Toronto, ON, M5S 1V4 Canada Phone: 1 416 946 5429 Email: [email protected] Abstract Background: Family violence (including intimate partner violence/domestic violence, child abuse, and elder abuse) is a hidden pandemic happening alongside COVID-19. The rates of family violence are rising fast, and women and children are disproportionately affected and vulnerable during this time. Objective: This study aims to provide a large-scale analysis of public discourse on family violence and the COVID-19 pandemic on Twitter. Methods: We analyzed over 1 million tweets related to family violence and COVID-19 from April 12 to July 16, 2020. We used the machine learning approach Latent Dirichlet Allocation and identified salient themes, topics, and representative tweets. Results: We extracted 9 themes from 1,015,874 tweets on family violence and the COVID-19 pandemic: (1) increased vulnerability: COVID-19 and family violence (eg, rising rates, increases in hotline calls, homicide); (2) types of family violence (eg, child abuse, domestic violence, sexual abuse); (3) forms of family violence (eg, physical aggression, coercive control); (4) risk factors linked to family violence (eg, alcohol abuse, financial constraints, guns, quarantine); (5) victims of family violence (eg, the LGBTQ [lesbian, gay, bisexual, transgender, and queer or questioning] community, women, women of color, children); (6) social services for family violence (eg, hotlines, social workers, confidential services, shelters, funding); (7) law enforcement response (eg, 911 calls, police arrest, protective orders, abuse reports); (8) social movements and awareness (eg, support victims, raise awareness); and (9) domestic violence±related news (eg, Tara Reade, Melissa DeRosa). Conclusions: This study overcomes limitations in the existing scholarship where data on the consequences of COVID-19 on family violence are lacking. We contribute to understanding family violence during the pandemic by providing surveillance via tweets. This is essential for identifying potentially useful policy programs that can offer targeted support for victims and survivors as we prepare for future outbreaks. (J Med Internet Res 2020;22(11):e24361) doi: 10.2196/24361 KEYWORDS Twitter; family violence; COVID-19; machine learning; big data; infodemiology; infoveillance COVID-19 a pandemic on March 11, 2020. To effectively Introduction control the spread of the disease, many countries have adopted As seen in the case of Ebola, epidemics increase the rates of rigorous measures to limit mobility, such as social distancing, domestic violence [1]. The World Health Organization declared stay-at-home orders (sheltering in place), closure of nonessential http://www.jmir.org/2020/11/e24361/ J Med Internet Res 2020 | vol. 22 | iss. 11 | e24361 | p. 1 (page number not for citation purposes) XSL·FO RenderX JOURNAL OF MEDICAL INTERNET RESEARCH Xue et al business, travel restrictions, and quarantine. Even though these medium, specifically the Internet, or in a population, with the measures are useful for infection control [2], they bring a series ultimate aim to inform public health and public policy. of negative social consequences, such as psychological stress Infodemiology data can be collected and analyzed in near real [3-5], unemployment [6], ageism [7], and increased rates of timeº (p 1). According to Eysenbach's framework, the violence against women and children [8-11]. Since these automated analysis of unstructured data related to family rigorous measures overlap with many of the intervention violence and COVID-19 is an application of an infoveillance strategies for family violence [2], they are likely to increase the study. Understanding public discussions can assist governments vulnerability of victims of family violence (including intimate and public health authorities in navigating the outbreak [32]. partner violence [IPV]/domestic violence, elder abuse, and child During the implementation of social isolation measures, social abuse), by increasing exposure to an exploitative relationship, media should be leveraged to raise public awareness and share reducing options for support [10], economic stress [12], and best practices (eg, bystander approaches, supportive statements, alcohol abuse [13,14]. For example, isolation limits social obtaining help on behalf of a survivor) [2], and provide support contact with families and social services, and thus may facilitate [33]. Twitter is a real-time network that allows users from across family violence and prevent victims from seeking help [15-17]. the globe to communicate via public and private messages, During the COVID-19 quarantine, the home becomes a organized chronologically on a given user©s account. Existing dangerous place for victims while individuals are living in forced studies have confirmed Twitter©s role in connecting practitioners close quarters [18]. In addition, mental health exacerbated by and clients [34-36]. Researchers have used Twitter data to social isolation increases the likelihood of locking victims of examine the nature of domestic violence [37-39]. A significant domestic violence in an unsafe home environment and increases number of studies describe Twitter hashtag #MeToo as a their vulnerability [19]. UNICEF [20] reports that school phenomenal tool for disclosing experiences of sexual closures increased child (sexual) abuse and neglect during the harassment, and more importantly, to ignite a widespread social Ebola epidemic. It is also important to note that child abuse and campaign or political protest on social media. Modrek and domestic violence are likely to co-occur when isolated at home Chakalov [40] examined tweets containing #MeToo in the [21,22]. During the COVID-19 pandemic, scholars have United States and supported the role of machine learning suggested that new forms of family violence may occur; for methods in understanding the widespread sexual assault example, abusers may threaten to infect their family members self-revelations on Twitter. Recently, Twitter has become a with the virus [23]. valuable source for understanding user-generated COVID-19 In many countries, the reported cases of and service needs content and activities in real time [41,42]. related to family violence dramatically increased since quarantine measures came into effect [18]. For example, calls Aim of the Study to domestic violence hotlines have risen by 25%, and the number There is a lack of data on the COVID-19 pandemic as it relates of Google searches for family violence±related help during the to family violence [43]. This study aims to provide a large-scale outbreak has been substantial [24]. According to National analysis of public discourse on family violence and COVID-19 Domestic Violence Hotline representatives in the United States, on Twitter using machine learning techniques to fill this gap. abusers are attempting to isolate victims from resources and The research questions are as follows: (1) what contents are unleashing more violence by enforcing COVID-19 social discussed relating to family violence and COVID-19? and (2) distancing measures [25]. In the United Kingdom, calls to the what themes are identified relating to family violence and Domestic Violence Helpline increase by 25% in the first week COVID-19? The study offers a new perspective on the impact, after the lockdown measures were implemented [26]. In China, risks factors, and continuing social support services during the domestic violence increased three times in Hubei Province pandemic for family violence. during the lockdown [27]. There was a 10.2% increase in domestic violence calls in the United States during the Methods COVID-19 pandemic [28]. These reports illustrate that existing COVID-19 intervention measures (eg, living in a closed space This study employed an observational design and followed the with abusers for a long period) may profoundly impact victims pipeline developed by the authors [44], including sampling, and survivors of family violence. According to Bradbury-Jones data collection, preprocessing of raw data, and data analysis. and Isham [8], ªdomestic violence rates are rising, and they are Sampling and Data Collection rising fastº (p 2047). Data on family violence during the Our COVID-19 data set used a list of COVID-19±relevant pandemic are still scarce [29], and there is a need for further hashtags as search terms to randomly collect tweets from Twitter research. between April 12 to July 16, 2020 [44] (Multimedia Appendix We cannot capture the impact of COVID-19 on family violence 1). Twitter Developer's Python code was used to access the without adequate surveillance [30]. Enhanced surveillance Twitter API to collect tweets. As shown in Figure 1, our data provides an understanding of the impact and risk factors set included